Landmark Detection
As part of our alignment algorithm for face recognition, we automatically detect anatomically defined points on the face across all poses, from frontal to profile faces. These include specific landmarks such as the center of the eye cavities, and the bridge, tip and base of the nose. While face alignment is our principal motivation behind developing accurate and precise landmark detectors, localization of certain landmarks, such as the eyes, can be an important component of other applications, such as iris-based recognition, as well.
Our landmark detectors are pose-specific. That is, we detect different sets of landmarks based on the head pose of the detected face. For frontal and semi-profile faces, we detect one set of landmarks, while for full-profile faces, we detect a second set of landmarks, since different points on the face are visible in each case. (See our software documentation for a detailed description of available landmarks.)
